CeTiO 2 N oxynitride perovskite: paramagnetic 14 N MAS NMR without paramagnetic shifts

: 14 N magic-angle spinning (MAS) nuclear magnetic resonance (NMR) spectra of diamagnetic LaTiO 2 N perovskite oxynitride and its paramagnetic counterpart CeTiO 2 N are presented. The latter, to the best of our knowledge, constitutes the ﬁ rst high-resolution 14 N MAS NMR spectrum collected from a paramagnetic solid material. The unpaired 4 f -electrons in CeTiO 2 N do not induce a paramagnetic 14 N NMR shift. This is remarkable given the direct Ce − N contacts in the structure for which ab initio calculations predict substantial Ce → 14 N contact shift interaction. The same effect is revealed with 14 N MAS NMR for SrWO 2 N (unpaired 5 d -electrons).


Introduction
LaTiO 2 N and other oxynitride semiconductors with the perovskite crystal structure exhibit favorable band gaps and visible-light-driven photocatalytic activity for water splitting reactions and hence constitute one of the most promising types of materials for this purpose [1][2][3][4][5][6][7][8][9][10]. Local O 2− /N 3− anion ordering in these systems is predicted to have significant implications on properties such as band gaps and was studied with X-ray, electron, and neutron diffraction, as well as first-principles calculations [11][12][13][14][15][16][17][18][19]. However, local O 2− /N 3− ordering occurs on too short length scales to be observed by diffraction techniques [14]. Moreover, O 2− and N 3− ions differ in charge, ionic radii, and coordination preferences and thus are not expected to occupy the same positions, as they are enforced in structures derived from diffraction measurements, which are biased by averaging. On the other hand, solid-state NMR probes nitrogen atoms directly, regardless of the chemical/ structural disorder. Therefore, in this work, we present 14 N magic-angle spinning (MAS) NMR spectra of LaTiO 2 N and its paramagnetic counterpart CeTiO 2 N, which may provide further insights into Ce/N/O arrangements in these structures. The paramagnetic nature of CeTiO 2 N was revealed by magnetic measurements resulting in a Weiss constant θ = −28 K and effective magnetic moment μ eff = 2.43 [14], in line with that expected for the free Ce 3+ ion (2.54), in contrast to Ti 3+ (1.73). To the best of our knowledge, this is the first-reported high-resolution 14 N MAS NMR spectrum of a solid paramagnetic material, considering previous attempts performed under static conditions and cryogenic temperatures. Schulman and Wyluda in 1962 reported 14 N continuous-wave (dispersion mode) NMR experiments on rare-earth nitrides (cubic structure): TbN and TmN. The observed signals were shifted significantly upfield with respect to the reference (liquid nitrogen) due to the hyperfine interaction with unpaired 4f electrons [20]. In 1968, Kuznietz reported 14 N spectra from ThN (upfield shift) and UN (downfield shift) [21,22]. In 2017, a static Fourier transform 14 N NMR spectrum of antiperovskite nitride Cr 3 GeN recorded at 4.2 K was reported revealing a 14 N quadrupolar coupling constant C Q = 0.3 MHz and a downfield shift [23].
Despite 99.6% natural abundance of the 14 N isotope, solid-state 14 N NMR studies of nitrogen-containing materials have rarely been reported. This is due to the spin I = 1 and a considerable nuclear quadrupole moment of 14 N, which result in signals that are severely broadened by quadrupolar interaction. However, as revealed by recent studies on diamagnetic Ta-and Nb-based oxynitrides and N-doped BaTiO 3 , 14 N MAS NMR spectra of these materials exhibit a single resonance without spinning sidebands, implying that quadrupolar interaction vanishes due to the high local symmetry of N 3− moieties [8,10,24,25]. This would only be expected for those with cubic crystal systems (BaNbO 2 N, BaTaO 2 N, BaTi(O,N) 3 ; Pm3m space group) [24,26], where the nitrogen anions occupy sites with octahedral symmetry (O h ). Nevertheless, although most of these oxynitride perovskites according to diffraction data develop space groups with lower than cubic symmetry, the appearance of 14 N MAS NMR spectra is essentially the same for all. This indicates that the local environments of nitrogen must be almost identical and that there are substantial deviations between the real local and averaged structures [25]. Fortuitous cancellation effects can also occur, in analogy to those that seem to be the case for ZnO ( 67 Zn C Q = 2.40 MHz) and ZnS ( 67 Zn C Q = 0.01 MHz), both having the same hexagonal (wurtzite) structure [27,28].
For paramagnetic systems, the total NMR shift can be expressed as δ = δ orb + δ con + δ pc (1) where the δ orb term corresponds to the orbital "chemical" shift, the sole shift contribution in diamagnetic systems, whereas the latter two terms arise due to the presence of unpaired electrons. The δ con is associated with the electronnucleus hyperfine coupling constant (HFCC) and denotes the effect of through-bond polarization called "contact shift", which is operative in close vicinity to paramagnetic ion. The δ pc term called "pseudocontact shift" (PCS) originates from electron-nucleus dipolar coupling and magnetic anisotropy of a paramagnetic center and is long range in nature (r −3 ). Paramagnetic lanthanide (Ln) ions can induce PCS effects for NMR-active nuclei at distances as long as 25−40 Å [29], which is exploited in biomolecular NMR to study large molecular systems [30][31][32]. On the other hand, solid-state NMR studies of materials incorporating paramagnetic lanthanide ions are scarce. At short distances between the nuclei of NMR interest and paramagnetic Ln 3+ ions in solids, both contact and pseudocontact mechanisms contribute to the NMR shift, which complicates data interpretation. 89 Y MAS NMR studies of Ln-substituted yttrium pyrochlores Y 2−x Ln x Sn 2 O 7 and Y 2−x Ln x Ti 2 O 7 revealed significant induced paramagnetic shifts of the 89 Y NMR signals [33]. Similar effects were observed for 27 Al resonances upon incorporation of Ce (or other paramagnetic lanthanides) in Y 3 Al 5 O 12 (YAG) [34][35][36], as well as for 43 Ca and 45 Sc NMR signals from CaSc 2 O 4 doped with Ce [37]. However, an understanding of the δ con /δ pc contributions could not be established. 17 O MAS NMR studies of paramagnetic lanthanide oxides Ln 2 O 3 revealed substantial 17 O shifts (hundreds to thousands ppm), with contributions from all three terms: δ orb , δ con , and δ pc [38,39]. In this work, we explore corresponding effects for nitrogen with 14 N MAS NMR on LaTiO 2 N and CeTiO 2 N.

Results and discussion
Let us first consider the 1 H MAS NMR spectrum of the LaTiO 2 N surface ( Figure 1a, black trace). The proton signal at 4.9 ppm corresponds to physisorbed H 2 O, whereas the remaining resonances originate from different types of bridging (>OH B ) and terminal hydroxyl groups (−OH T ). In contrast, surface proton signals from paramagnetic CeTiO 2 N (Figure 1a, red trace) are strongly affected by the presence of unpaired electrons of Ce 3+ ions and paramagnetic NMR interactions arising thereof [32]. The 14 N MAS NMR spectrum of LaTiO 2 N is shown in Figure 1b, black trace. The narrow signal at 270 ppm is consistent with the data reported for other diamagnetic oxynitride perovskites that are collected in Table 1. The 14 N MAS NMR spectrum of paramagnetic CeTiO 2 N is presented in Figure 1b, red trace. Despite being broader and slightly less symmetric, the 14 N signal from CeTiO 2 N exhibits no induced paramagnetic shift due to unpaired 4f electrons when compared to its LaTiO 2 N counterpart and other diamagnetic oxynitride perovskites in Table 1. This is remarkable given the direct Ce−N contacts in the network and strong paramagnetic effects observed for surface protons ( Figure 1a). This counterintuitive result could be explained by the cancellation of orbital and paramagnetic shift contributions with opposite signs, as it is partially the case for 17 O NMR shifts in Sm 2 O 3 [39]. However, based on the 14 N NMR data in Table 1 for 10 diamagnetic oxynitride perovskites incorporating a variety of metal ions, it is safe to assume that δ orb does not change for CeTiO 2 N as well, so the induced paramagnetic shift has to be zero, or δ con and δ pc have to cancel each other out.
To explore these scenarios, the induced paramagnetic 14 N NMR shifts were evaluated in terms of electron paramagnetic resonance (EPR) parameters by applying the formalism of Moon and Patchkovskii [42] and Vaara [43][44][45][46][47][48] to the cluster model derived from the CeTiO 2 N crystal lattice. The calculations of 14 N hyperfine coupling tensors and the electronic g-tensor of Ce were performed with ab initio quantum chemistry methods: domain-based local pair natural orbital coupled cluster singlets and doublets (DLPNO-CCSD) [49] and multireference perturbation theory (CASSCF/NEVPT2) [50], respectively. Although this theoretical approach employed an approximation of the averaged structure derived from diffraction measurements, the molecular orbital theory is helpful because it provides an upper/lower boundary of the range of 14 N HFCCs expected at the Ce−N distances representative for the CeTiO 2 N crystal lattice and insight into relative magnitudes of contact and PCS contributions. The ab initio approach proposed herein for paramagnetic NMR shifts prediction constitutes a potent addition to recent applications of post-Hartree-Fock methods for the NMR shift calculations in diamagnetic systems, providing accuracy beyond standard density functional theory (DFT) (see the SI) [49,[51][52][53][54].
The predicted 14 N HFCCs for nitrogen positions at the distances of 2.1 and 2.5 Å from Ce in the CeTiO 2 N model are −2.0 and −0.9 MHz, respectively (see Table S5 and negative spin densities on nitrogen in Figure 2). These values are in the same range as the experimental estimates of −2.2 and −0.6 MHz for 17  The calculated induced 14 N paramagnetic contact shifts for nitrogen atoms at the distances of 2.1 and 2.5 Å from Ce 3+ ions are −498 and −213 ppm, respectively (Table S8 in the SI), whereas magnitudes of pseudocontact shifts are <30 ppm and diminish to <10 ppm already at distances ⩾3.3 Å. Therefore, long-range effects from distant paramagnetic centers are unlikely to counteract substantial contact shifts of close Ce−N contacts, given that the next nearest Ce neighbors are >4.5 Å away.
However, as can be seen in Figure 2a, the negative spin density region is associated with the 3d-orbital of one of the Ti atoms, in contrast to the mostly positive region observed  on the Ce 4f-orbital. This indicates that the nitrogen atom involved in the bond to this particular titanium atom (site N 2 in the model; Figure S7) may experience cancellation effects of the two. Noteworthily, the calculated induced paramagnetic shift (δ con + δ pc ) for site N 2 (see Table S8) of −1.4 ppm (without effects of other Ce 3+ centers) is close to the experimental paramagnetic shift of −2 ppm (peak maximum). This cancellation effect of spin densities associated with Ce 4f and Ti 3d orbitals under some particular Ce−N−Ti occupational arrangement could potentially also explain why there is no paramagnetic 14 N NMR shift in CeTiO 2 N compared with paramagnetic rare-earth nitrides.
To shed light on local symmetry effects, the 14 N MAS NMR spectrum of NdTiO 2 N was collected. NdTiO 2 N is not fully stoichiometric (predicted composition of NdTiO 2.17 N 0.83 ) and exhibits statistical anion distribution resulting in O/N occupational disorder in contrast to other perovskite oxynitrides [55][56][57]. Therefore, the high local symmetry of nitrogen environments in these systems is not expected to be fully satisfied in NdTiO 2 N. And this is the case, as the 14 N MAS NMR signal of NdTiO 2 N was broadened (nearly) beyond detection (Figure 1b, gray trace), which we attribute to occupational disorder combined with paramagnetic effects. This suggests that both the precise O/N settlement in the lattice and its arrangement with respect to Ce and Ti are the origin of the unexpected 14 N MAS NMR spectrum of CeTiO 2 N.
To inspect for the presence of potential impurities occurring in the CeTiO 2 N sample, transmission electron microscopy was employed. We could not discern the existence of any additional phases and the selected area electron diffraction pattern was identified to belong to CeTiO 2 N and to be consistent with powder X-ray diffraction patterns (see Figure S4 in the SI). Noteworthily, titanium nitride exhibits an approximately 90 ppm higher 14 N NMR shift compared with those observed herein [28,58,59]. Moreover, to test the hypothesis of amorphous TiO x /N y impurities, X-ray photoelectron spectroscopy (XPS) analyses were performed for LaTiO 2 N and CeTiO 2 N samples, as well as for the reference amorphous TiO x /N y film prepared by atomic layer deposition [60]. The latter thin film was obtained by depositing first an amorphous TiO x layer, followed by ammonolysis at 623 K. These reaction conditions are known to provide amorphous TiO x /N y layers. The XPS characteristics of the amorphous TiO x /N y reference film are clearly distinct from those of LaTiO 2 N and CeTiO 2 N (see Figure S5). This indicates that the collected 14 N MAS NMR signals do not originate from a potential amorphous side phase TiO x /N y .
In addition, we have synthesized SrWO 2 N and recorded its 14 N MAS NMR spectrum. SrWO 2 N has a cubic structure (Pm3m; Table 1) with tungsten ions being in the 5d 1 electron configuration [61] and exhibits metallic character, which makes it challenging for MAS NMR. The sharp 14 N NMR signal (Figure 1b; blue trace) is consistent with the octahedral symmetry of the nitrogen ions in the structure. The 14 N shift of 269 ppm is basically the same as for the 10 diamagnetic oxynitrides presented in Table 1. This renders the 14 N shift unsusceptible to the type, oxidation state, and electronic configuration of the cations present in these structures. It is worth noting that hexagonal phases of ZrN, InN, and GaN metal nitrides exhibit similar shifts [59].

Conclusions
To conclude, no paramagnetic 14 N NMR shift was observed for CeTiO 2 N with respect to LaTiO 2 N and other diamagnetic oxynitride perovskites. This constitutes the first highresolution 14 N MAS NMR spectrum collected from paramagnetic solid material. Ab initio calculations predict 14 N hyperfine coupling constants to be similar to those estimated for oxygen in lanthanide oxides by 17 O MAS NMR experiments, where effects due to paramagnetic Ln 3+ ions are substantial. Whereas the invariance of the 14 N chemical shift in diamagnetic oxynitrides to the type and oxidation state of the cations involved is quite odd, the absence of any NMR effect either from 4f or from 5d electrons in CeTiO 2 N and SrWO 2 N is truly remarkable. Since these features cannot be captured in diffraction-derived, averaged crystal structures, more work is needed to gain insight into local environments in these unconventional materials.  respectively) with a Bruker Avance III spectrometer equipped with a 1.3 mm MAS probehead and employing a MAS rate of 60.00 kHz (43.00 kHz for SrWO 2 N due to its metallic character and difficulties with sample spinning). 1 H NMR acquisitions were performed with a rotorsynchronized, double-adiabatic spin-echo sequence with a 90°excitation pulse of 1.1 μs, followed by two 50.0 μs tanh/tan short high-power adiabatic pulses [62,63] with a 5 MHz frequency sweep. All pulses operated at a nutation frequency of 210 kHz. 256 signal transients with a 5 s relaxation delay were accumulated. 14 N MAS NMR spectra were collected using a 3.0 μs 90°excitation pulse and 65,536 scans collected per sample using a 1 s relaxation delay. The spectrum of SrWO 2 N was recorded with the Hahn-echo sequence. 1 H shifts were referenced to neat tetramethylsilane (TMS) at 0 ppm, whereas 14 N shifts to solid NH 4 Cl at 0 ppm (−342.4 ppm with respect to nitromethane).
Author contributions: All the authors have accepted responsibility for the entire content of this submitted manuscript and approved submission. Research funding: None declared. Conflict of interest statement: The authors declare no conflicts of interest regarding this article.